Predictive analytics tool offers clinicians actionable EHR data

NorthShore University HealthSystem has a long history of predictive modeling to identify patients at high risk for adverse outcomes like cardiac arrest and readmissions. But until recently the Evanston, Ill.-based health system didn't have an easy way for clinicians to act on the information. The electronic health record generated multiple risk scores for different conditions that were difficult to track.

Interested in giving providers more actionable information, NorthShore developed the Clinical Analytics Predictive Engine, a tool in the EHR that assigns each patient a single risk score tied to multiple predictive models.

The tool, which went live at NorthShore in late September, evaluates in real-time patients' risk factors for various adverse outcomes and generates an overall score in the EHR based on the data. It then offers a checklist of interventions the clinician should follow based on the risk score.

“It's moving from predictive to prescriptive modeling,” said Chad Konchak, assistant vice president of health information technology at NorthShore.

Known as CAPE, the tool took a year to develop in collaboration with the University of Chicago and Epic Systems Corp., NorthShore's EHR vendor. It currently addresses risk for cardiac arrest, 30-day readmission and mortality.

The tool continuously takes in critical clinical information about patients including their vital signs, results from labs and comorbidities. It also processes patients' prior utilization and admission history to gauge the risk of readmission.

Based on the results, patients are assigned a score displayed in the EHR through colors. Green means the patient isn't at a high risk for any adverse outcome, yellow indicates the patient shows some signs of risk and red indicates a patient is at high risk.

For the yellow and red displays, the EHR alerts the appropriate clinician about what intervention should be taken based on the condition the patient is at risk for. Different members of the care team are alerted based on what's needed.

For example, if a patient is at risk for cardiac arrest, red indicates the patient should be transferred to the intensive-care unit for closer monitoring. Yellow means additional lab tests should be ordered and more frequent vital sign checks should be conducted.

Case managers are alerted for patients at risk for a readmission. NorthShore has a program specifically for patients at risk for readmission, using care coordinators who ensure patients have medications before discharge and follow up with them at home.

“It's prescribing interventions to all the different people across the care continuum,” Konchak said.

NorthShore tested the tool for about three months with clinical staff to ensure it was safe. The system has preliminary data that show the tool has resulted in a decline of cardiac arrest patients and reductions in mortality.

“It's working; it's providing value to physicians,” said Dr. Nirav Shah, an infectious-disease physician at NorthShore who helped design the tool.

Initially, physicians were hesitant, Shah said. They feel they know their patients best and were on the fence about trusting a machine. But as data came in showing the approach helped prevent adverse outcomes, buy-in improved, he said.

NorthShore is tracking staff's compliance in following the tool's suggested interventions, but Konchak stresses that the tool isn't meant to take away physician autonomy.

The tool will also be a way for NorthShore to quickly test new interventions. The tool can randomly assign similar patients an intervention while withholding that option for other patients. NorthShore staff can then compare outcomes between the groups. And because the tool automatically determines who gets an intervention and who doesn't, additional resources or staff time aren't required.

“This cuts down on how much time it takes to test out whether an intervention works or not and this can have profound implications to be able to conduct quality initiatives,” Shah said.